Covariance of preference and performance on normal and novel hosts in a locally monophagous and locally polyphagous butterfly population

Covariance between preference and performance was quantified for Papilio glaucus strains derived from a locally monophagous Florida population and a locally polyphagous Ohio population. I used two-choice assays to assess relative host preferences of mothers for plant species that represent reciprocal normal and novel hosts for each population (i.e., Liriodendron tulipifera and Magnolia vrginiana) and a split-brood design to quantify relative performance of their progeny on each host. Covariance between preference and proxies of performance was detected independently within each population, which is a level of genetic structure at which such covariance has rarely been documented. These results support the hypothesis that preference-performance covariance can exist in populations that have no obvious internal host-associated structure. In the Ohio strain, all proxies of performance (larval duration, pupal mass, relative growth rate, and survival) were significantly correlated with relative preference for the normal host, L. tulipifera. In the Florida strain, however, only pupal mass was correlated with maternal preference, and this correlation was not in the direction expected. Progeny that attained the heaviest mass were derived from mothers that preferred L. tulipifera, the locally rare host. The nature of the preference-performance links was not in the manner predicted by conventional optimal oviposition theory, whereby host-associated tradeoffs have been considered an implicit element. Relative performance was congruent across hosts, regardless of whether mothers preferred L. tulipifera, M. virginiana, or neither host. After considering possible genetic and nongenetic explanations that could account for preference-performance covariance in P. glaucus, I conclude that this covariance has a genetic basis. Likely, multiple genetic control mechanisms (e.g., pleiotropy and co-adaptation) integrate at the level of different trait combinations and/or a particular trait combination to generate observed patterns.